Method and system for detecting and analyzing local voltage distribution of aluminum-silicon eutectic bath

Abstract

The application belongs to the technical field of data processing, and provides a eutectoid groove of aluminum-silicon alloy local voltage distribution detection and analysis method and system, the method comprises the following steps: obtaining key voltage data of the eutectoid groove of aluminum-silicon alloy; performing multidimensional preprocessing and physical field coupling modeling reconstruction on the key voltage data to obtain a standardized voltage distribution matrix; based on the standardized voltage distribution matrix and the key voltage data, performing dynamic feature adaptive extraction to obtain multidimensional feature data; according to the key voltage data, the standardized voltage distribution matrix and the multidimensional feature data, performing multilevel voltage anomaly detection on the eutectoid groove of aluminum-silicon alloy to obtain an anomaly detection result. The scheme realizes deep fusion and intelligent analysis of multidimensional data, greatly improves detection accuracy, improves anomaly detection accuracy through multidimensional field coupling modeling and dynamic feature extraction, thereby improving detection reliability, and full-automatic analysis reduces manual operation errors, and detection intelligence is higher.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell. Background Technology

[0002] In the field of aluminum electrolysis production, the eutectoid cell for aluminum-silicon alloys is the core equipment for preparing aluminum-silicon alloys, and its operating status directly determines product quality, production efficiency, and energy consumption costs. Based on the Hall-Herrouhl process, this equipment uses molten cryolite-alumina as the electrolyte and initiates an electrolytic reaction with direct current at a high temperature of 940-980℃ to generate liquid aluminum-silicon alloys. However, due to the complexity of the eutectoid cell's structure, the strong magnetic field interference generated by the high current, the temperature gradient within the cell, and uneven anode consumption, the local voltage distribution within the cell is prone to abnormalities. These abnormalities manifest as a sudden increase in voltage gradient at the edges, local voltage distortion caused by magnetic field distortion, and resistance-increasing voltage fluctuations caused by material accumulation. If these abnormalities are not controlled in a timely manner, they can lead to a chain reaction of failures such as local overheating, electrode erosion, and decreased electrolysis efficiency, and even major safety accidents such as cell leakage.

[0003] Currently, the local voltage detection technology for aluminum-silicon alloy eutectoid cells still has many shortcomings, making it difficult to meet the high-precision, real-time, and intelligent requirements of industrial production. Specific problems are as follows: On the one hand, traditional detection methods mostly focus on measuring the overall voltage of the eutectoid cell. Affected by high-temperature environments and strong electromagnetic interference, the sampling accuracy is insufficient to meet practical high-precision requirements, and it cannot capture microvolt-level local voltage fluctuations. On the other hand, existing detection schemes only perform simple filtering and threshold comparison on the collected voltage data, resulting in a high false alarm rate and frequent invalid shutdown checks, severely impacting production continuity. Furthermore, in existing detection schemes, the detection equipment and data analysis system are independent, and data transmission relies on manual copying or low-speed industrial buses. The cycle from data acquisition to anomaly detection is long, leading to delays in handling sudden faults, making it difficult to meet actual needs in terms of detection timeliness. Moreover, the detection process has a low level of intelligence and insufficient reliability.

[0004] It is evident that current local voltage detection schemes for aluminum-silicon alloy eutectoid cells suffer from technical problems such as low detection accuracy, poor real-time performance, low level of intelligence, and insufficient detection reliability. Summary of the Invention

[0005] This invention provides a method and system for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell, which solves the defects of current aluminum-silicon alloy eutectoid cell local voltage detection schemes, such as low detection accuracy, poor real-time performance, low level of intelligence, and insufficient detection reliability.

[0006] On one hand, the present invention provides a method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell, comprising: Obtain key voltage data for the eutectoid cell of aluminum-silicon alloy; The key voltage data are preprocessed in multiple dimensions and reconstructed by physical field coupling modeling to obtain a standardized voltage distribution matrix; Based on the standardized voltage distribution matrix and the key voltage data, dynamic feature adaptive extraction is performed to obtain multi-dimensional feature data; Based on the key voltage data, the standardized voltage distribution matrix, and the multidimensional feature data, multi-level voltage anomaly detection is performed on the aluminum-silicon alloy eutectoid cell to obtain the anomaly detection results.

[0007] The method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell provided by the present invention obtains key voltage data of the aluminum-silicon alloy eutectoid cell, including: Multiple voltage detection points are evenly distributed at each discharge port in the lower feeding area of ​​the aluminum-silicon alloy eutectoid cell; Acquire at least one set of local voltage data for each voltage detection point, and use all local voltage data as key voltage data.

[0008] According to the method for detecting and analyzing the local voltage distribution in the eutectoid cell of aluminum-silicon alloy provided by the present invention, the key voltage data is preprocessed in multiple dimensions and reconstructed by physical field coupling modeling to obtain a standardized voltage distribution matrix, including: The key voltage data is filtered and denoised, temperature compensated, and neighborhood signal collaboratively corrected to obtain multi-dimensional preprocessed key voltage data. Based on the multidimensional preprocessed key voltage data, a standardized voltage distribution matrix is ​​obtained through physical field coupling modeling and reconstruction.

[0009] The method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell provided by the present invention includes temperature compensation processing of the key voltage data, comprising: Establish a voltage-temperature coupling correction model; Obtain the measured temperature data corresponding to the key voltage data, and perform temperature compensation processing on the key voltage data based on the measured temperature data and the voltage-temperature coupling correction model.

[0010] The method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell according to the present invention includes performing neighborhood signal collaborative correction on the key voltage data, comprising: Correlation analysis is performed on multiple adjacent voltage signals within the same detection area in the key voltage data to calculate the signal similarity coefficient; Abnormal voltage signals with a similarity coefficient lower than a preset similarity coefficient threshold are extracted, and the abnormal voltage signals are corrected based on a pre-built basic parameter library.

[0011] According to the method for detecting and analyzing the local voltage distribution in the eutectoid cell of aluminum-silicon alloy provided by the present invention, based on the key voltage data after multidimensional preprocessing, a standardized voltage distribution matrix is ​​obtained through physical field coupling modeling and reconstruction, including: A three-dimensional voltage distribution calculation model is established based on the multi-physics coupling effect of electric field, magnetic field and temperature field in the eutectoid cell of aluminum-silicon alloy. Based on the three-dimensional voltage distribution calculation model and the multi-dimensional preprocessed key voltage data, the initial voltage distribution matrix is ​​reconstructed. The initial voltage distribution matrix is ​​standardized to obtain the standardized voltage distribution matrix.

[0012] According to the method for detecting and analyzing the local voltage distribution in the eutectoid cell of aluminum-silicon alloy provided by the present invention, based on the standardized voltage distribution matrix and the key voltage data, dynamic feature adaptive extraction is performed to obtain multi-dimensional feature data, including: Based on the standardized voltage distribution matrix, spatial dimension features are extracted to obtain spatial domain features; Based on the standardized voltage distribution matrix and the key voltage data, time-domain features are extracted to obtain time-domain features; An optimization algorithm combining principal component analysis and mutual information entropy is used to dynamically and adaptively optimize the spatial and temporal features to obtain multidimensional feature data.

[0013] According to the method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell provided by the present invention, based on the key voltage data, the standardized voltage distribution matrix, and the multidimensional feature data, multi-level voltage anomaly detection is performed on the aluminum-silicon alloy eutectoid cell to obtain anomaly detection results, including: Based on the voltage distribution range under various operating conditions in the pre-built basic parameter library, the voltage dynamic threshold is determined. Based on the voltage dynamic threshold, anomaly detection is performed on the standardized voltage distribution matrix to obtain the first-level detection result. A voltage trend prediction model is established, the key voltage data is input into the voltage trend prediction model to obtain the voltage trend prediction result, and anomaly detection is performed based on the voltage trend prediction result to obtain the secondary detection result; A distribution pattern detection model is established, and the multidimensional feature data is input into the distribution pattern detection model to detect whether the voltage distribution pattern is abnormal, so as to obtain a three-level detection result. Based on the first-level detection results, the second-level detection results, and the third-level detection results, anomaly detection results are obtained through correlation anomaly verification.

[0014] The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to the present invention further includes: Based on the anomaly detection results, the root cause of the voltage anomaly in the aluminum-silicon alloy eutectoid cell was identified, and the anomaly risk level was determined. Based on the root cause of the voltage anomaly and the anomaly risk level, optimization suggestion information is generated; The anomaly detection results, root causes of voltage anomalies, anomaly risk levels, and optimization suggestions are used as the report content to generate a detection analysis report.

[0015] On the other hand, the present invention also provides a system for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell, comprising: The acquisition module is used to obtain key voltage data of the aluminum-silicon alloy eutectoid cell; The preprocessing module is used to perform multi-dimensional preprocessing and physical field coupling modeling and reconstruction on the key voltage data to obtain a standardized voltage distribution matrix; The extraction module is used to perform dynamic adaptive feature extraction based on the standardized voltage distribution matrix and the key voltage data to obtain multi-dimensional feature data; The detection module is used to perform multi-level voltage anomaly detection on the aluminum-silicon alloy eutectoid cell based on the key voltage data, the standardized voltage distribution matrix, and the multi-dimensional feature data, and obtain the anomaly detection results.

[0016] The method and system for detecting and analyzing local voltage distribution in eutectoid cells of aluminum-silicon alloys provided by this invention accurately acquires key voltage data, reconstructs a standardized voltage distribution matrix through multi-dimensional preprocessing and physical field coupling modeling, and combines dynamic feature adaptive extraction and multi-level anomaly detection to achieve deep fusion and intelligent analysis of multi-dimensional data. This significantly improves detection accuracy and effectively solves the shortcomings of traditional detection schemes in capturing subtle faults. Through multi-physics field coupling modeling and dynamic feature extraction, the accuracy of anomaly detection is improved, and the false alarm rate and false negative rate are significantly reduced, reducing ineffective maintenance intervention and thus improving detection reliability. The fully automated analysis reduces human operation errors, resulting in a higher level of detection intelligence. The stable voltage distribution can also improve the purity and quality of aluminum-silicon alloy products, thereby helping the aluminum electrolysis industry achieve intelligent and energy-saving upgrades. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the voltage acquisition principle in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the aluminum-silicon alloy eutectoid cell local voltage distribution detection and analysis system provided in this embodiment of the invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] The following is combined with Figures 1 to 3 This invention describes the detailed scheme of the method and system for detecting and analyzing the local voltage distribution in the eutectoid cell of aluminum-silicon alloy provided in the embodiments of the present invention.

[0021] like Figure 1 As shown in the embodiment of the present invention, the method for detecting and analyzing the local voltage distribution in an aluminum-silicon alloy eutectoid cell mainly includes the following steps: Step 110: Obtain key voltage data for the aluminum-silicon alloy eutectoid cell.

[0022] In this embodiment, multiple voltage detection points can be arranged at the key detection locations of the aluminum-silicon alloy eutectoid cell, and an array-type dot matrix detection method can be used to achieve more comprehensive voltage detection.

[0023] Step 120: Perform multi-dimensional preprocessing and physical field coupling modeling and reconstruction on the key voltage data to obtain the standardized voltage distribution matrix.

[0024] Understandably, multi-dimensional preprocessing can filter out noise points in key voltage data, improve data quality, and combined with physical field coupling modeling and reconstruction, a more accurate and comprehensive representation of voltage distribution can be obtained.

[0025] Step 130: Based on the standardized voltage distribution matrix and key voltage data, perform dynamic adaptive feature extraction to obtain multidimensional feature data.

[0026] Understandably, through dynamic feature adaptive extraction, features can be accurately extracted from multiple dimensions, thereby providing effective data for subsequent anomaly detection.

[0027] Step 140: Based on key voltage data, standardized voltage distribution matrix and multidimensional feature data, perform multi-level voltage anomaly detection on the aluminum-silicon alloy eutectoid cell to obtain anomaly detection results.

[0028] In this embodiment, voltage anomaly detection through multi-level voltage anomaly detection can achieve voltage anomaly analysis from multiple levels, thereby improving the accuracy and reliability of anomaly detection results.

[0029] In one embodiment, obtaining key voltage data for an aluminum-silicon alloy eutectoid cell specifically includes: First, multiple voltage detection points are evenly distributed at each discharge port in the lower feeding area of ​​the aluminum-silicon alloy eutectoid cell.

[0030] Understandably, the feeding area at the bottom of the aluminum-silicon alloy eutectoid cell is the core area where alumina material enters. The uniformity of material distribution in this area directly affects the sufficiency of the electrolytic reaction. If too much material is fed, it will cause material accumulation, which will increase local resistance and cause the voltage to rise. If the material is not fed enough, it will cause the electrolyte to be exposed, which will aggravate electrode corrosion and may also cause local overheating, which will also lead to abnormal voltage.

[0031] Accordingly, this embodiment selects six feeding areas as key detection areas, mainly based on the following two points: First, the six feeding areas are evenly distributed at the bottom of the electrolytic cell, which can fully cover the core reaction area of ​​the electrolytic cell and avoid missing anomalies due to too few collection points; Second, the voltage changes in the feeding areas are most sensitive to the material state and electrolyte stability, and the voltage signal in these areas can be used to quickly capture subtle anomalies in the production process, providing accurate basis for real-time control of feeding parameters.

[0032] In practical applications, such as Figure 2 As shown, one voltage detection point can be set in each of the six feeding areas at the bottom of the electrolytic cell. The voltage detection points must be precisely aligned with the core reaction area below the feeding port to ensure stable contact between the sensor probe and the electrolyte. The sensor and transmission line must be resistant to high temperature, corrosion, and electromagnetic interference to adapt to the harsh environment of high temperature, strong magnetic field, and corrosive gas around the electrolytic cell. The six voltage detection points must be symmetrically and evenly arranged to ensure the representativeness and comparability of the detection data.

[0033] Then, at least one set of local voltage data is obtained for each voltage detection point, and all local voltage data are used as key voltage data.

[0034] In practical applications, each feeding area can utilize a high-precision voltage sensor for data acquisition. This sensor, with its probe-like structure, contacts the electrolyte or electrodes within the electrolytic cell to collect local voltage data in real time. For example... Figure 2 As shown, the collected local voltage data is transmitted as an analog signal to the acquisition module via a shielded cable. The acquisition module converts the analog signal into a digital signal, which is then transmitted to the tank control machine via an industrial bus or wireless communication.

[0035] During the pre-installation preparation stage, the electrolytic cell needs to be shut down and cooled, and debris and accumulated material in the lower part of the cell's feeding area need to be cleaned. The position coordinates of the six feeding ports need to be checked, and the installation center position of each voltage detection point needs to be marked to ensure that the spacing between each point is uniform. Installation materials such as high-temperature resistant voltage sensors, shielded cables, insulating fasteners, and high-temperature resistant sealant need to be prepared.

[0036] During the sampling point drilling and insulation stage, a special high-temperature resistant tool can be used to drill a mounting hole at the marked installation location. The size of the mounting hole should match the sensor probe. Apply high-temperature resistant insulating sealant to the inner wall of the mounting hole and insert an insulating sleeve to prevent the sensor from contacting the metal part of the tank and causing a short circuit. This also enhances the sealing of the installation and prevents electrolyte leakage.

[0037] During the sensor installation and fixing stage, the voltage sensor probe can be passed through the insulating sleeve and inserted into the electrolytic cell. The probe insertion depth should be controlled at 5-10cm below the feed port to ensure that the probe can stably contact the electrolyte and avoid damage from the impact of the feed. The sensor is fixed to the cell wall by the stainless steel fixing flange and the fixing bolts are tightened to ensure that the sensor is firmly installed and there is no looseness.

[0038] During the line layout and protection phase, the output lines of each voltage sensor can be connected with shielded cables. The cable lines are laid along a dedicated cable tray on the outside of the tray to avoid crossing with high-voltage lines and reduce electromagnetic interference. High-temperature resistant and waterproof joints are used to seal the cable joints, and the entire line is wrapped with a high-temperature resistant protective sleeve to prevent damage to the line from high temperature and corrosive gases.

[0039] During the debugging and calibration phase, after installation, the voltage sensors at the six voltage detection points were debugged to check whether the communication between the voltage sensors and the acquisition module was normal and whether the acquired voltage signals were stable. Each voltage sensor was calibrated using a standard voltage source to ensure the accuracy of the detection data. Finally, no-load and load tests were conducted to verify the operational stability of the entire detection system.

[0040] In one embodiment, key voltage data undergoes multidimensional preprocessing and physical field coupling modeling and reconstruction to obtain a standardized voltage distribution matrix, specifically including: First, the key voltage data is filtered and denoised, temperature compensated, and neighboring signal collaboratively corrected to obtain multi-dimensional preprocessed key voltage data.

[0041] In this embodiment, a wavelet threshold denoising algorithm can be used to decompose the voltage time series of a single voltage detection point in the key voltage data. The number of decomposition layers can be set to 5. The db4 wavelet basis function is selected, and the adaptive threshold is dynamically adjusted based on the signal variance and noise intensity to remove high-frequency noise components and retain effective signal features.

[0042] In one specific implementation, temperature compensation processing is performed on key voltage data, specifically including: The first step is to establish a voltage-temperature coupling correction model.

[0043] Understandably, based on the electrolytic reaction characteristics and multi-physics coupling principle of the aluminum-silicon alloy eutectoid cell, and taking the normal operating temperature range of the eutectoid cell as a basis, firstly, voltage sample data under different temperature gradients can be collected experimentally. Specifically, in the laboratory, the high-temperature environment of the cell is simulated, and parameters such as electrolyte composition and current intensity are controlled to be stable. Only the temperature variable is changed, and the local voltage value corresponding to each temperature point is recorded simultaneously to form a sample dataset containing the relationship between temperature and voltage.

[0044] Subsequently, combining the physical law of electrolyte conductivity changing with temperature, a coupled correction model was constructed using a fusion algorithm of multiple linear regression and BP neural network. With temperature as the input variable and voltage correction as the output variable, the basic correlation trend between temperature and voltage was first fitted by linear regression, and then the nonlinear deviation was learned by BP neural network. The model parameters were optimized through iterative training, and finally a voltage and temperature coupled correction model that can accurately reflect the law of temperature influence on voltage was obtained.

[0045] The second step is to obtain the measured temperature data corresponding to the key voltage data, and then perform temperature compensation processing on the key voltage data based on the measured temperature data and the voltage-temperature coupling correction model.

[0046] In practical applications, PT100 temperature sensors can be simultaneously deployed at the voltage detection points in the aluminum-silicon alloy eutectoid cell, maintaining spatial consistency with the voltage sensors to ensure accurate correspondence between the collected temperature data and key voltage data. During voltage data acquisition, the measured temperature data at each voltage detection point is acquired simultaneously, and combined with the corresponding raw key voltage data and timestamps, a three-dimensional data set of voltage-temperature-time is constructed.

[0047] The measured temperature data is input into the established voltage-temperature coupling correction model to calculate the voltage correction amount under that temperature condition. Following the principle that the corrected voltage equals the sum of the original voltage and the correction amount, point-by-point compensation is performed on the key voltage data at each voltage detection point to eliminate voltage measurement deviations caused by temperature gradients. This ensures that voltage data under different temperature environments have uniform comparability, thus providing accurate data support for the subsequent construction of a standardized voltage distribution matrix.

[0048] In one specific implementation, neighborhood signal collaborative correction is performed on key voltage data, specifically including: The first step is to perform correlation analysis on multiple adjacent voltage signals within the same detection area in the key voltage data and calculate the signal similarity coefficient.

[0049] In this embodiment, 3-5 spatially adjacent voltage detection points can be considered as a neighborhood signal group to ensure coverage of local areas under the same physical environment and operating conditions. Based on the Pearson correlation coefficient algorithm, pairwise correlation analysis is performed on the time series of each voltage signal within each neighborhood group to calculate the linear correlation between any two adjacent voltage signals, obtaining the pairwise similarity coefficient, with a value range of [-1, 1]. Simultaneously, a dynamic time warping algorithm is used to compensate for the insufficient adaptation of the Pearson coefficient to nonlinear trend signals, calculating the morphological similarity of the signal waveforms. Finally, a weighted fusion is performed to obtain the signal similarity coefficient, thus comprehensively reflecting the consistency of voltage signals within the neighborhood.

[0050] The second step is to extract abnormal voltage signals whose signal similarity coefficient is lower than the preset similarity coefficient threshold, and to correct the abnormal voltage signals based on the pre-built basic parameter library.

[0051] In practical applications, the signal similarity coefficients of all neighboring signal groups can be iterated to filter out abnormal voltage signals whose similarity coefficients are lower than a preset similarity coefficient threshold. Then, a pre-built basic parameter library is called, which contains standard voltage distribution models for each detection location under different operating conditions, neighboring signal correlation patterns, and historical amendment examples.

[0052] First, by matching the timestamp and location coordinates of the abnormal voltage signal with the standard signal characteristics of the corresponding operating condition in the basic parameter library, the direction of the abnormal deviation is determined, such as being too high, too low, or fluctuating abnormally. Then, based on the mean and trend slope of the normal signal in the neighborhood and the voltage correlation parameters in the standard model, a combination of interpolation algorithm and trend fitting is used to correct each sampling point of the abnormal voltage signal point by point. This ensures that the corrected signal not only conforms to the cooperative change law of the neighborhood signal, but also matches the standard characteristics in the basic parameter library, thus eliminating abnormal deviations caused by environmental interference and instantaneous sensor failures.

[0053] Then, based on the key voltage data after multidimensional preprocessing, the standardized voltage distribution matrix is ​​obtained by physical field coupling modeling and reconstruction.

[0054] In one specific implementation, based on the multidimensional preprocessed key voltage data, a standardized voltage distribution matrix is ​​obtained through physical field coupling modeling and reconstruction, specifically including: The first step is to establish a three-dimensional voltage distribution calculation model based on the multi-physics coupling effect of the electric field, magnetic field and temperature field of the aluminum-silicon alloy eutectoid cell.

[0055] In this embodiment, the high-temperature operating conditions and electrolytic reaction characteristics of the aluminum-silicon alloy eutectoid cell can be combined. Based on Ohm's law and Kirchhoff's voltage law, the multi-physics coupling mechanism of electric field, magnetic field and temperature field can be comprehensively considered. The three-dimensional mesh of the cell can be divided by the finite element analysis method, and parameters such as electrolyte composition, electrode structure and electrode spacing can be incorporated into the model. Then, a three-dimensional voltage distribution calculation model that can accurately reflect the influence of each physical field on the local voltage can be constructed.

[0056] The second step is to reconstruct the initial voltage distribution matrix based on the three-dimensional voltage distribution calculation model and the multi-dimensional preprocessed key voltage data.

[0057] In practical applications, the key voltage data after multidimensional preprocessing can be used as boundary conditions to input the three-dimensional voltage distribution calculation model. The voltage values ​​of the areas where no measurement points are set up can be completed by solving the model, forming an initial voltage distribution matrix covering the entire key area of ​​the eutectoid cell. The matrix elements correspond to the local voltage data of each spatial location, realizing a comprehensive characterization of the voltage distribution.

[0058] The third step is to standardize the initial voltage distribution matrix to obtain the standardized voltage distribution matrix.

[0059] Specifically, the Z-score normalization algorithm can be used to normalize all voltage data in the initial voltage distribution matrix based on the voltage mean and voltage standard deviation of the initial voltage distribution matrix. This eliminates the differences in reference voltage and the influence of dimensions in different regions, so that the standardized voltage data are in a uniform and comparable range. Finally, a standardized voltage distribution matrix that can support subsequent feature extraction and anomaly detection is obtained.

[0060] In one embodiment, based on the standardized voltage distribution matrix and key voltage data, dynamic adaptive feature extraction is performed to obtain multi-dimensional feature data, specifically including: On the one hand, based on the standardized voltage distribution matrix, spatial dimension features are extracted to obtain spatial domain features.

[0061] In this embodiment, a standardized voltage distribution matrix is ​​used as the core data carrier. The focus is on the spatial distribution law of voltage in the eutectoid cell. Parameters reflecting the uniformity and difference characteristics of voltage in the region are extracted. Specifically, the spatial gradient of the matrix, regional variance, voltage difference between the edge and center regions, and the average and peak voltage values ​​of different feeding regions can be calculated to obtain spatial features that can comprehensively characterize the spatial distribution state of voltage.

[0062] On the other hand, based on the standardized voltage distribution matrix and key voltage data, time-dimensional features are extracted to obtain time-domain features.

[0063] In this embodiment, the time-series voltage data of each measuring point in the standardized voltage distribution matrix can be combined with the original key voltage data. The time-domain features can be extracted by sliding window, and the mean, variance, kurtosis, peak factor, voltage change trend slope, fluctuation frequency and duration of the voltage signal at each voltage detection point can be calculated. At the same time, the voltage change amplitude and rise rate of sudden situations such as anode effect can be captured to obtain time-domain features that can characterize the dynamic change law of voltage.

[0064] Finally, an optimization algorithm combining principal component analysis and mutual information entropy was used to dynamically and adaptively optimize the spatial and temporal features to obtain multidimensional feature data.

[0065] Specifically, spatial and temporal features can be integrated into a high-dimensional original feature vector. First, principal component analysis is used to linearly reduce the dimensionality of the high-dimensional features, retaining principal components with a cumulative contribution rate of over 95% and eliminating linearly correlated redundant features. Then, mutual information entropy is used to calculate the nonlinear correlation between the remaining features, eliminating redundant feature pairs with mutual information entropy higher than 0.9. By dynamically and adaptively adjusting the feature selection threshold, multidimensional feature data with optimized dimensionality and prominent core information is finally obtained, providing efficient input for subsequent anomaly detection.

[0066] In one embodiment, based on key voltage data, a standardized voltage distribution matrix, and multidimensional feature data, multi-level voltage anomaly detection is performed on the aluminum-silicon alloy eutectoid cell to obtain anomaly detection results, specifically including: On the one hand, based on the voltage distribution range under various operating conditions in the pre-built basic parameter library, the voltage dynamic threshold is determined. Based on the voltage dynamic threshold, anomaly detection is performed on the standardized voltage distribution matrix to obtain the first-level detection result.

[0067] Specifically, a pre-built basic parameter library can be retrieved, which also contains voltage distribution range data under different operating conditions. Combined with the real-time operating parameters of the aluminum-silicon alloy eutectoid cell, a dynamic threshold algorithm is used to calculate the corresponding voltage dynamic threshold (including upper and lower limits) for each operating condition. Each voltage data point in the standardized voltage distribution matrix is ​​compared with the voltage dynamic threshold one by one, and abnormal voltage points exceeding the threshold range are marked, forming a first-level detection result that includes the location, value, and over-limit amplitude of the abnormal voltage point.

[0068] On the other hand, a voltage trend prediction model is established, key voltage data is input into the voltage trend prediction model to obtain voltage trend prediction results, and anomaly detection is performed based on the voltage trend prediction results to obtain secondary detection results.

[0069] In this embodiment, a voltage trend prediction model can be constructed based on a long short-term memory network. Historical key voltage data is used as training samples to optimize model parameters, enabling the model to accurately fit the voltage time series variation pattern. Then, real-time collected key voltage data can be input into the voltage trend prediction model according to a fixed time window to predict the voltage change trend over a preset time period. The deviation rate between the predicted and actual values ​​is calculated. If the deviation rate exceeds a set threshold and the duration meets the standard, it can be marked as an abnormal trend, forming a secondary detection result.

[0070] On the other hand, a distribution pattern detection model is established, and multi-dimensional feature data is input into the distribution pattern detection model to detect whether the voltage distribution pattern is abnormal, so as to obtain the three-level detection results.

[0071] Specifically, a distribution pattern detection model can be constructed using an algorithm that combines isolated forests and support vector machines. The training set uses multi-dimensional feature data, including feature samples from normal and typical abnormal scenarios such as material accumulation and electrode erosion. The model is trained to learn the feature differences of different voltage distribution patterns. The multi-dimensional feature data to be detected is then input into the trained distribution pattern detection model. A model voting mechanism determines whether the voltage distribution pattern matches an abnormal pattern, outputting the abnormality matching degree and the corresponding abnormality type, forming a three-level detection result.

[0072] Finally, based on the results of the first-level, second-level, and third-level detections, the anomaly detection results are obtained through correlation anomaly verification.

[0073] In practical applications, a spatial correlation map and temporal synchronization analysis model for measurement points can be constructed to integrate the detection results of Level 1, Level 2, and Level 3, and analyze the spatial correlation and temporal synchronization between anomalies at different levels. Cross-validation is performed on anomalies that meet the correlation conditions to eliminate isolated false anomalies, accurately locate the center position, impact range, and core anomaly type of the anomaly area, and finally output anomaly detection results containing anomaly level, cause analysis, and location information.

[0074] In some embodiments, if the anomaly detection results show that there is an abnormal voltage condition, a voltage anomaly warning message can be generated in a timely manner, and the anomaly warning can be given in a variety of ways such as on-site warning, SMS reminder, and message pop-up, so as to ensure that the abnormal situation can be handled in a timely manner.

[0075] In one embodiment, the above-mentioned method for detecting and analyzing the local voltage distribution in the eutectoid cell of aluminum-silicon alloy may further include: First, based on the anomaly detection results, the root cause of the voltage anomaly in the aluminum-silicon alloy eutectoid cell was identified, and the anomaly risk level was determined.

[0076] In this embodiment, the anomaly detection results can be combined with the basic parameter database and production condition data to trace the root cause through a causal analysis model. If the voltage in the abnormal area is high and the corresponding material feed exceeds the standard, it is determined to be material accumulation; if the voltage fluctuation is accompanied by an increase in temperature, it is determined to be electrode erosion based on the electrode consumption record. At the same time, based on the scope of the anomaly's impact, the magnitude of the voltage deviation, and the rate of fault development, the risk level is divided into three levels: low, medium, and high. For example, a localized, single-point, slight anomaly is considered low risk, while a multi-area synchronous anomaly with a sudden voltage surge is considered high risk.

[0077] Then, based on the root cause of the voltage anomaly and the level of anomaly risk, optimization suggestions are generated.

[0078] In practical applications, based on the identified root causes and risk levels of anomalies, a process optimization knowledge base containing historical amendment examples and parameter adjustment rules is invoked to generate optimization suggestions. For example, if the anomaly is related to material accumulation, which is at a medium-risk level, the output will be a reduction ratio for material feeding and a solution for adjusting the feeding frequency; if the anomaly is related to electrode erosion, which is at a high-risk level, the output will be suggestions for electrode position correction and replacement cycle; if the anomaly is related to uneven current distribution, which is at a low-risk level, the output will be suggestions for fine-tuning current density parameters. The optimization suggestions must clearly define the adjustment objectives, operational steps, and expected results to ensure that they can directly guide on-site operation and maintenance.

[0079] Finally, the abnormal detection results, root causes of voltage abnormalities, abnormal risk levels, and optimization suggestions are included in the report to generate a detection analysis report.

[0080] Specifically, it can integrate anomaly detection results, root causes of voltage anomalies, anomaly risk levels, and optimization suggestions, presenting them in a structured manner according to a standardized template. The detection analysis report includes five main modules: summary, detailed detection data, anomaly analysis, optimization suggestions, and conclusions. It can automatically embed real-time data charts, such as voltage distribution heatmaps and trend curves, and supports exporting in a set format. At the same time, the detection analysis report can be synchronized to the production management system for operators and technicians to review and make decisions.

[0081] Based on the same general inventive concept, this invention also protects a local voltage distribution detection and analysis system for an aluminum-silicon alloy eutectoid cell. The local voltage distribution detection and analysis system for an aluminum-silicon alloy eutectoid cell provided by this invention will be described below. The local voltage distribution detection and analysis system for an aluminum-silicon alloy eutectoid cell described below can be referred to in correspondence with the local voltage distribution detection and analysis method for an aluminum-silicon alloy eutectoid cell described above.

[0082] like Figure 3 As shown in the embodiment of the present invention, the local voltage distribution detection and analysis system for the eutectoid cell of aluminum-silicon alloy specifically includes: The acquisition module 210 is used to acquire key voltage data of the aluminum-silicon alloy eutectoid cell.

[0083] The preprocessing module 220 is used to perform multi-dimensional preprocessing and physical field coupling modeling and reconstruction on key voltage data to obtain a standardized voltage distribution matrix.

[0084] The extraction module 230 is used to perform dynamic adaptive feature extraction based on the standardized voltage distribution matrix and key voltage data to obtain multidimensional feature data.

[0085] The detection module 240 is used to perform multi-level voltage anomaly detection on the aluminum-silicon alloy eutectoid cell based on key voltage data, standardized voltage distribution matrix and multi-dimensional feature data, and obtain anomaly detection results.

[0086] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments of the relevant methods, and will not be elaborated further here.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell, characterized in that, include: Obtain key voltage data for the eutectoid cell of aluminum-silicon alloy; The key voltage data are preprocessed in multiple dimensions and reconstructed by physical field coupling modeling to obtain a standardized voltage distribution matrix; Based on the standardized voltage distribution matrix and the key voltage data, dynamic feature adaptive extraction is performed to obtain multi-dimensional feature data; Based on the key voltage data, the standardized voltage distribution matrix, and the multidimensional feature data, multi-level voltage anomaly detection is performed on the aluminum-silicon alloy eutectoid cell to obtain the anomaly detection results.

2. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Obtain key voltage data for the eutectoid cell of aluminum-silicon alloy, including: Multiple voltage detection points are evenly distributed at each discharge port in the lower feeding area of ​​the aluminum-silicon alloy eutectoid cell; Acquire at least one set of local voltage data for each voltage detection point, and use all local voltage data as key voltage data.

3. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, The key voltage data undergoes multidimensional preprocessing and physical field coupling modeling and reconstruction to obtain a standardized voltage distribution matrix, including: The key voltage data is filtered and denoised, temperature compensated, and neighborhood signal collaboratively corrected to obtain multi-dimensional preprocessed key voltage data. Based on the multidimensional preprocessed key voltage data, a standardized voltage distribution matrix is ​​obtained through physical field coupling modeling and reconstruction.

4. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 3, characterized in that, Temperature compensation processing is performed on the key voltage data, including: Establish a voltage-temperature coupling correction model; Obtain the measured temperature data corresponding to the key voltage data, and perform temperature compensation processing on the key voltage data based on the measured temperature data and the voltage-temperature coupling correction model.

5. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 3, characterized in that, The key voltage data is subjected to neighborhood signal collaborative correction, including: Correlation analysis is performed on multiple adjacent voltage signals within the same detection area in the key voltage data to calculate the signal similarity coefficient; Abnormal voltage signals with a similarity coefficient lower than a preset similarity coefficient threshold are extracted, and the abnormal voltage signals are corrected based on a pre-built basic parameter library.

6. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 3, characterized in that, Based on the multidimensional preprocessed key voltage data, a standardized voltage distribution matrix is ​​obtained through physical field coupling modeling and reconstruction, including: A three-dimensional voltage distribution calculation model is established based on the multi-physics coupling effect of electric field, magnetic field and temperature field in the eutectoid cell of aluminum-silicon alloy. Based on the three-dimensional voltage distribution calculation model and the multi-dimensional preprocessed key voltage data, the initial voltage distribution matrix is ​​reconstructed. The initial voltage distribution matrix is ​​standardized to obtain the standardized voltage distribution matrix.

7. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Based on the standardized voltage distribution matrix and the key voltage data, dynamic adaptive feature extraction is performed to obtain multi-dimensional feature data, including: Based on the standardized voltage distribution matrix, spatial dimension features are extracted to obtain spatial domain features; Based on the standardized voltage distribution matrix and the key voltage data, time-domain features are extracted to obtain time-domain features; An optimization algorithm combining principal component analysis and mutual information entropy is used to dynamically and adaptively optimize the spatial and temporal features to obtain multidimensional feature data.

8. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Based on the key voltage data, the standardized voltage distribution matrix, and the multidimensional feature data, multi-level voltage anomaly detection is performed on the aluminum-silicon alloy eutectoid cell to obtain anomaly detection results, including: Based on the voltage distribution range under various operating conditions in the pre-built basic parameter library, the voltage dynamic threshold is determined. Based on the voltage dynamic threshold, anomaly detection is performed on the standardized voltage distribution matrix to obtain the first-level detection result. A voltage trend prediction model is established, the key voltage data is input into the voltage trend prediction model to obtain the voltage trend prediction result, and anomaly detection is performed based on the voltage trend prediction result to obtain the secondary detection result; A distribution pattern detection model is established, and the multidimensional feature data is input into the distribution pattern detection model to detect whether the voltage distribution pattern is abnormal, so as to obtain a three-level detection result. Based on the first-level detection results, the second-level detection results, and the third-level detection results, anomaly detection results are obtained through correlation anomaly verification.

9. The method for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell according to any one of claims 1 to 8, characterized in that, The method further includes: Based on the anomaly detection results, the root cause of the voltage anomaly in the aluminum-silicon alloy eutectoid cell was identified, and the anomaly risk level was determined. Based on the root cause of the voltage anomaly and the anomaly risk level, optimization suggestion information is generated; The anomaly detection results, root causes of voltage anomalies, anomaly risk levels, and optimization suggestions are used as report content to generate a detection analysis report.

10. A system for detecting and analyzing local voltage distribution in an aluminum-silicon alloy eutectoid cell, characterized in that, include: The acquisition module is used to acquire key voltage data of the aluminum-silicon alloy eutectoid cell; The preprocessing module is used to perform multi-dimensional preprocessing and physical field coupling modeling and reconstruction on the key voltage data to obtain a standardized voltage distribution matrix; The extraction module is used to perform dynamic adaptive feature extraction based on the standardized voltage distribution matrix and the key voltage data to obtain multi-dimensional feature data; The detection module is used to perform multi-level voltage anomaly detection on the aluminum-silicon alloy eutectoid cell based on the key voltage data, the standardized voltage distribution matrix, and the multi-dimensional feature data, and obtain the anomaly detection results.

Images